Harvey Open-Sources RL Env for M&A Due Diligence
harvey · x · 2026-07-18
Harvey has open-sourced 10 RL environments for M&A due diligence. Agents must conduct research within a Virtual Data Room (VDR) and write due diligence memos for the deals.
The largest environment features a context of 80 million tokens and 1,000 scoring criteria. The author points out that these tasks are particularly difficult for frontier agents:
- Long context: A single research thread might require processing hundreds or thousands of documents in the data room.
- High exhaustiveness: It requires finding deal-breaking clauses, much like finding a needle in a haystack.
- Data scarcity: Real due diligence data is privately held by transactional law firms, meaning models have rarely seen it before.
The post also emphasizes that M&A due diligence itself is high-value knowledge work: the global M&A market is roughly $4 trillion, with about 1% spent specifically on due diligence.
More from coding & agent
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22
- Claude Code skill uses 10 Markdown rules to make outputs ADHD-friendly — alex_verem · 2026-07-22